Energy Recycling in Photonics — Three Things Called “Recycling” | Uday Sonawane

[0:00]I'm UD Sonowane. Energy recycling photonic hardware sounds like one thing. It is at least three at three different layers of the machine and they are not interchangeable. Here is how to tell which one a paper is actually claiming. First, the phrase search for energy recycling photonics and you get three separate ideas wearing one label. Reusing the pump light that drives a device. Reusing the photons themselves before anything measures them. And recovering waste heat downstream which is not optical at all. Those sit at different layers. They are measured against different baselines and a result in one says nothing about the others. Three questions for any photonic

[0:35]Recycling claim. One, what exactly is being reused? Pump power photons or heat. Two, is it reused before or after detection? Because detection destroys the photon and that is a hard physical wall, not an engineering problem. Three, what is the number compared against the laser alone or the whole system including the electronics? Start with the strongest published example. A team at Stanford built an optical amplifier on thin film lithium neobate published in Nature this January. Normally the pump light that drives an amplifier passes through once and whatever is left leaves. Here the pump is generated inside a resonator so it circulates and passes the same material again and

[1:10]Again. Nothing is reclaimed from waste. The same light is simply put back to work before it can escape. The numbers from the paper more than 17 dB of gain drawing under 200 mwatt. 95% of the input is converted into the pump that does the work and the noise stays flat and near the quantum limit across 110 nanometers of bandwidth which matters because the cheap way to get gain is to accept more noise and they did not. So what did going round again by two things and the paper states both against conventional single pass designs the effective pump power rises by nearly an order of magnitude same laser roughly 10

[1:44]Times the useful pump because the light keeps passing the gain medium instead of leaving. And against previous demonstrations, the power needed to reach this gain drops by about an order of magnitude. That is the whole claim. And it is a claim about pump light. Now, the part that does not survive the headline. First, this is an amplifier, not a processor. It makes a signal bigger. It does not perform a computation. And nothing here says a photonic computer got more efficient. Second, and this is the wall in optical computing. The photon is destroyed when you detect it. You can recirculate light before measurement never after. And

[2:17]Third, when I searched for energy recycling in photonics, most of what came back was not about reclaiming light at all. It was about spending less power tuning phase shifters and moving data. Real work, different claim. The other two layers do exist. In particle acceleration, there is a proposal to loop the light left over after it has pushed an electron back around for the next cycle. So, photons that would have become heat get another pass. And in photonic neural networks, there are recirculating mesh architectures that reuse the optical path before anything is detected. Both are the same word at a different layer. Both are early. And I

[2:48]Am deliberately not putting numbers on either one. Scored. What is reused? Pump light in the one result with hard numbers. Before or after detection. Before always and after is not an engineering gap. It is physics. Compared against what? Single pass designs and prior amplifiers both named in the paper. So be excited that resonant recirculation turned roughly a 10-fold power saving into a real device on a real platform. Be skeptical of the phrase itself because it will be attached to tuning power results and to heat recovery, which are different claims wearing the same word. Your turn. Next time you meet an energy recycling photonix claim, ask what is physically

[3:26]Being reused, whether it happens before detection, and what baseline the saving is measured against. If the answer is waste heat, it is a data center result, not a computing one. Three things called recycling.